Could Deep Blue play a smarter AI ?

[NapoléonPremier]

(...)What would happen? The human would say, 'I can make that jump.' And would probably hurt himself trying to jump to the other platform. Heh, heh, I could just imagine the frustration of the scientist doing the study as subject after subject disregarded the warning (and the broken bodies below) and still thought: I can make that jump.

That's it. That's the difference. You see it everywhere. Nothing is impossible. Nothing is unprovable. We can do anything; we just need time.

So, there you have it. Apparently, to be intelligent like us; sentient like us; an animal (or computer), would have to be arrogant, crazy and self-deluded to the point where it thinks it can do anything.

Funny ain't it?

This is a very interesting story. The spider behavior and "intelligence", and even that of some more primitive forms of life like bacteria is indeed fascinating (btw, I'm not sure I got this right : did the spider evaluate the right rod by climbing each one of the three, or did it "evaluate" the right one from the distance, which would be even more amazing).
And your hypothesis about human behavior in the same situation rings very true...
But yes, you're right, this is precisely this hubris, this will to power, this idea that nothing is impossible which, although causing many catastrophies on the way, allowed us to go from not much more than apes to where we are today in less than 10000 years... So I say kudos to all of us men (and some women too ), and especially to the arrogant, risk-taking, crazy MF ones...

 

[jar2574]

Deep Blue was not brute force.

Deep Blue applies brute force aplenty, but the "intelligence" is the old-fashioned kind.

??? ???

Part of Deep Blue's success comes from brute force. The articles on the site support that. Articles from many other sites do as well.

Your first statement is incorrect. Deep Blue does use brute force, it just does it in a more "intelligent" manner than in the past.

If you don't know what you're talking about, it's better not to speak up.

If you do know what you're talking about then you are not expressing yourself properly. But as usual you are ripping on others in the process.

 

[NapoléonPremier]

cognito ergo sum

That's "cogito", latin meaning "I think"

Edit: Here's one: http://wwwcs.uni-paderborn.de/~IPCCC//
They use lots of big words! lol! I have no idea what "heuristics" means!

"Heuristics" is a method for solving problems based on experiences as opposed to pure logic. In chess AI, one of the methods used to accelerate the computing is called "null-move heuristics" (I think somebody posted a link to that earlier in the thread).

 

[NapoléonPremier]

Some would say that all we're doing is simulating sentience.

That's an old science-fiction classic, how do we know we're not just androids programmed to resemble humans ?
Cf Philip K. Dick's "Do androids dream of electric sheeps ?" and "Blade Runner", the film made after it.

 

[NapoléonPremier]

I don't personally think we're anymore special than other animals, or even plants or bacteria for that matter.

I agree, you're probably not...

 

[NapoléonPremier]

I understood what you meant Zombie - I was just hoping to also hear what the other side of the opposition had in mind when supporting the "brute force" of Deep Blue. Because in this aspect it is self evident that Deep Blue wasn't using such a brute force; otherwise, they wouldn't need some chess grandmasters to TRAIN a computer how to calculate a tree to the end.

Although Deep Blue is "groomed" with "lessons learned" from its huge database of games and openings, the core of it is brute force - that is a decision tree-based algorithm (or algorithms), and the ability to evaluate hundreds of millions of possible moves every second.
On the other hand, the Civ AI for instance is based on a "set of rules" approach, which explains why it can do so many dumb things. As mentioned before, the complexities of the game would make a "brute force/decision tree" approach impossible within current computing capabilities.
But I thought we already covered all that in detail in the beginning of the thread. Maybe that's another weakness of us humans : we never really learn...

 

[NapoléonPremier]

Maybe we should be programming a computer to problem solve like a spider, insect, or heck even other mammals; seriously. It would be a lot more practical. Maybe it would even a better opponent, since humans do get bogged down in psychological hangups that other intelligences don't have.

It's true that humans have flaws that animals don't have, but all in all it seems pretty obvious that the efficiency of human beings as problem solving machines is far superior to anything else that exists on this planet. Otherwise it would be spiders, or cows, sitting at their spider-made computers, who would be discussing these subjects right now (and maybe being smarter than us ).

 

[5cats]

That's "cogito", latin meaning "I think"

"Heuristics" is a method for solving problems based on experiences as opposed to pure logic. In chess AI, one of the methods used to accelerate the computing is called "null-move heuristics" (I think somebody posted a link to that earlier in the thread).

Doh! Latin is far too weird for me

Thanks, I was guessing heuristics meant something like that.

All the chess software runs on brute force. The suble differences between them are in the areas of:
How far ahead does it look?
How does it "score" each position?
How fast can it go?
How does it manage time?
What does it avoid?

By avoid I mean stupid moves that are obviosly bad and shouldn't be 'explored' which would waste precious time. Where does it abandon that logic tree? A Rook sacrifice looks bad, but might win you the game!

Philip K. Dick's "Do androids dream of electric sheeps ?" and "Blade Runner", the film made after it.

PK Dick was a great writer! He had several themes running through his books & stories, and a unique style all his own.

(btw, I'm not sure I got this right : did the spider evaluate the right rod by climbing each one of the three, or did it "evaluate" the right one from the distance, which would be even more amazing).

It's my understanding this was done from a distance. ie trial and error wasn't used, the spider figured out which was the correct rod much better than chance would allow. Spider intelligence is quite different from ours, we understand it even less than our own!

I guess I'll check in to the Nut-House too, since I think I'm a cat & all...

 

[warpus]

Brute force in programming means exactly what you just said. As opposed to having a superior heuristic allowing one to search only some branches and ignore inferior branches. Or better yet, to work without any tree in the first place!

It seems obvious to me that Deep Blue uses a brute-force decision-tree algorithm coupled with pre-programmed knowledge from older games (played by grandmasters), allowing it to prune branches much more effectively.

That's still considered a brute-force approach, though, it's just fairly efficient.

I personnally agree that, to my knowlege, there is no "scientific" evidence that can back the existence of "souls" (whatever that means). However, my point is that if you, as an individual, want to know if you really are a spiritual being and not just a bunch of carbon cells excited by electrical currents, this is something which can be achieved without too much difficulty, as long as you're motivated. With pretty good certainty. But not to "prove" it to the others. For yourself. Which is all that really matters in the end.

Yeah, I can easily close my eyes, meditate for a bit, and then have a funny feeling in my stomach.. "Holy crap, my body is full of thetans!". Is that what you're talking about?

You can convince yourself of anything if you really want to. It doesn't make anything true, however.

I completely agree that the number of people believing something proves nothing. However, India is not some primitive tribe from the Pacific believing in some "Great Frog God". Today, India is a very advanced nation, counting some of the best mathematicians and computer programmers in the world. Yet they've been holding the same beliefs for more than 5000 years, far longer than christianity. And a lot of Westerners, starting from the 19th century and before, who have studied these beliefs have been strongly shaken by them. And these beliefs are not just some "do's" and "don'ts" like what christianity has mostly become. They're a very complex and very subtle metaphysical view of life, man and the universe.

So? A large number of people believing something proves nothing, you agree with that. But now you're saying that if these people believe that something for thousands of years, then it DOES prove something? Bollocks!

If a billion people believed, for 10,000 years, that humans can fly of their own accord, it wouldn't make it any more true.

 

[atreas]

It seems obvious to me that Deep Blue uses a brute-force decision-tree algorithm coupled with pre-programmed knowledge from older games (played by grandmasters), allowing it to prune branches much more effectively.

That's still considered a brute-force approach, though, it's just fairly efficient.

Again, "seems obvious" doesn't include any knowledge about how these things are done. For example, try to explain the following commentary of Kasparov himself in his last book:

"It seems it is beyond the power of the program to remove a rook from an open file".

What this phrase means in "no chess argo" is that there is a rule inside the program that evaluates the changing positions (like "a rook in open files is worth more"). Now the program creates the tree, TO SOME DEPTH, and evaluates each variation according to this and many other rules (such as piece mobility, material, weaknesses, mating possibilities, etc.). As long as some other factor doesn't change (and in slowly manovering game they rarely do) the computer can't change its evaluation - just because it is not using brute force, to examine the trees at full depth in all directions. Of course it is calculating more and better (more accurately) than any human, but this is just an aspect of its power.

The database has three roles: 1) to teach an opening guide, 2) to teach an endgame guide, and 3) to provide a way to tune up the relative significance of the evaluation parameters. In chess middlegame it's not very practical to search in the database for exactly similar games played before.

 

[warpus]

Again, "seems obvious" doesn't include any knowledge about how these things are done. For example, try to explain the following commentary of Kasparov himself in his last book:

"It seems it is beyond the power of the program to remove a rook from an open file".

What this phrase means in "no chess argo" is that there is a rule inside the program that evaluates the changing positions (like "a rook in open files is worth more"). Now the program creates the tree, TO SOME DEPTH, and evaluates each variation according to this and many other rules (such as piece mobility, material, weaknesses, mating possibilities, etc.). As long as some other factor doesn't change (and in slowly manovering game they rarely do) the computer can't change its evaluation - just because it is not using brute force, to examine the trees at full depth in all directions. Of course it is calculating more and better (more accurately) than any human, but this is just an aspect of its power.

The database has three roles: 1) to teach an opening guide, 2) to teach an endgame guide, and 3) to provide a way to tune up the relative significance of the evaluation parameters. In chess middlegame it's not very practical to search in the database for exactly similar games played before.

The thing is that the algorithm is BASED on a brute-force decision-tree approach. It is supplemented with heuristic algorithms (these are used to prune branches as well as quantifying the "goodness" of branches) which are very efficient - they're based on data obtained from previous games, played by grandmasters.

So yeah, while it might not examine every single node in the tree to determine the next move, a decision-tree algorithm is at the heart of Deep Blue.

 

[Brighteye]

Sorry everyone if you can't remember this part of the thread. It was a way back.

Nope. Your real point is that souls don't exist. But you're not arguing it in a very effective manner. You're trying to use science to discredit the belief in souls. That won't work.

If souls do exist then they don't operate under your scientific laws. Just because you make an object that you think "ought to be intelligent," does not mean that a soul "should" or "will" enter the object.

The very conception of souls is incompatible with the notion that they "will" enter into an object just because it is identical to the brain. People who believe in souls don't think they only exist within the brain.

I gave two very clear options. Either souls exist, and have no effect on the physical world, or they don't exist. This is the whole point of my thought experiment. If souls do exist they don't operate under scientific laws, but the laws remain laws governing nature. This is the foundation of science. If you do not accept the laws that science has so far made, then there are plenty of things, such as light bulbs, cars etc. that you need to explain with your theory of everything. If souls are part of nature (the physical world) and allow complex brains to have free will, escaping determinism, then they follow physical laws, and will be present in every such brain, however it was created. This is because if they are part of the physical world they follow the principle (that applies in the real world) that identical objects will behave in the same way, and this includes either having free will or not.

Ah, here is your real point.

I thought my real point was the previous one? Or are all my points real?

And science has certainly limited the domain of God through its discoveries.

But unfortunately this claim is still untestible because we cannot create an identical brain. Even more unfortunately for your "thought experiment", even if we could create one, an identical brain would not need to contain a soul for people's belief in souls to remain intact.

Because, as I said, I was not proving that souls do not exist, but merely defining their influence on the world if they do. Therefore of course I won't stop people believing in them. I do not see why it is unfortunate for my thought experiment, unless you are being peculiarly sensitive and see me as someone who is spitefully attempting to destroy religion through my thought experiment. This is not the case.

The irony here is that my "irrelevant" paragraphs were merely responding to your statement that: "If we build a brain that is identical with another one then how we built it does not matter, because it is now identical, and so there is no way for it to remember the order in which the parts were built."

My statements were only meant as responses to correct your illogical point of view. Obviously, no brain remembers how it was put together. It's unfortunate that you failed to find my comments relevant within that context. I'm not going to argue that they are relevant to the overall discussion, because they were only meant as responses to a sentence that was illogical and really quite silly.

What is illogical and silly about my sentence? Please explain. I postulated an identical object, and you blather on about how brains develop differently even if they develop from the same cell and in the same medium. Given that Ihad postulated an identical object, any comments about development are irrelevant, because it's .... IDENTICAL. Have I said that enough times yet? It's a thought experiment, and the practicality of making these identical objects is also not relevant.

Your point that souls don't exist will not be proven in a thought experiment about identical brains. I should have cut to the heart of the matter earlier and shown how your "thought experiment" couldn't prove the existence of souls one way or the other. My apologies.

You ended up finding my comments about the difficulty of creating an identical human brain irrelevant. But at first you responded to them. And you responded in ways that made it seem as though you did not appreciate the complexity of the human brain and the enourmous task that would go into replicating it. I'm sorry for that brief diversion, and will try to remain focused on this "thought experiment" that you have proposed instead of the difficulties involved in creating an actual experiment resembling it.

My point was not that souls do not exist. I didn't state this clearly at first, because I didn't realise you hadn't understood. Having stated it clearly I fail to see why you still do not comprehend. This is not what I was saying.
I have ample knowledge about the problems of actually making a brain, but as I have said, the whole point of a thought experiment is that you temporarily ignore these problems in order to understand a point that the experiment is making. Not only did you not understand the concept of a thought experiment, but you also did not understand the point I was making with it.

 

[Brighteye]

Absolutely not. I never suggested that souls make us intelligent. I never suggested that the argument 'souls = intelligence' was a valid scientific argument either.

Here is the basic structure of this confusing thread as I see it.

You: "we can create intelligence"
Others: "we can't create intelligence because souls are needed"
You: "souls aren't needed, because of X"
Me: "X will not convince Others or other believers that souls do not exist."
You: "Are you saying that the Others raised a valid argument?"
Me: "Nope, I'm saying that X will not convince Others that they are wrong."

I argued that the "thought experiment" which tried to disprove the existence of the invisible monkeys/souls was not going to convince people who believed in monkeys/souls that they do not exist. The "thought experiment" was worthless if its goal was to convince people that monkeys/souls do not exist.

'...if its goal was to convince people that monkeys/souls do not exist.'
However, the point of the thought experiment fits neatly into your summary because it is meant to demonstrate that souls are irrelevant to creating intelligence either because they aren't necessary, or because there will inevitably be one in an AI that we create.

 

[Brighteye]

But that's not an identical object. What we know is that the "soul" (assuming there is one, again) enters the WHOLE BODY at a certain point during conception/gestation. An identically grown brain is not the same object as a whole body. So even with your own concepts, there's no guarantee at all.

Sure, good point. But if you're granting me the theoretical ability to make an identical brain, let me change my thought experiment and make an identical body to go with the brain. Make the whole thing that the soul inhabits. The point it demonstrates is the same.

In maths, it's true that we can establish that some things can never be proven within a certain finite logical system - again, that's Godel. And even more simply, one of the fundamental basis of maths and logic are axioms, which by definition cannot be proven.
But you're being sophistic, or maybe forgot what the original debate was about : not if some things are impossible in general. But if we can be absolutely certain that there are practical results that will always be impossible for science to achieve (traveling faster than light, change lead into gold etc...). And to that again, I say no, we can never be sure in advance.

Maybe I did lose track of the point... it's easy to do in 20 or so pages.

This coming from a guy who was wisely explaining a few posts before how logical debates can so easily turn personal... Don't worry, I'm not offended. Cause if I was, I would kind of prove I am indeed stupid. But if say nothing, then I implicitly agree I am stupid too ?

Sorry! Yes, I was a bit ebullient. But then again, calling one line of thought stupid is a little different from calling a person stupid. Everyone makes mistakes/slips up/ doesn't concentrate from time to time.

 

[Brighteye]

...And I guarantee you won't regret the trip (which doesn't mean you'll have to shave your head and walk around in the streets chanting "hare hare Krishna" ).

I just read about Krishna. Apparently he has a girlfriend in every city. No wonder they think he's a God! Managing to keep that situation going deserves respect, if not worship.

 

[jar2574]

I gave two very clear options. Either souls exist, and have no effect on the physical world, or they don't exist. This is the whole point of my thought experiment.

Then your thought experiment did not prove anything about intelligence, the topic of discussion in the thread.
I thought you were trying to say that souls did not exist, because I thought you were trying to say that your thought experiment would prove that souls have nothing to do with intelligence.

No one has proven that human intelligence is solely a product of the physical world. No one has proven that souls do not exist. So under your thought experiment souls could exist and have no effect on the physical world, but have an effect on intelligence. That's why your conclusion that a replica of a human brain would have to be intelligent is not necessarily true. Your conclusion is only true if we assume that souls do not exist. That's why I assumed that you were saying that souls do not exist.

 

[Pawel]

Heh, heh. You mean all my letters on 'Ether Exists' published in the Penthouse Forums wouldn't look good on my resume?

No, that is actually extra credit!

Sure, the behaviour of a large number of electrons isn't random - it can be predicted with a probability wave. However, the behaviour of 1 individual electron is nondeterministic. I hope we can agree to that.

Sure, agreeing is nice! And you're basically right, even though there is actually no 'probability wave'. If you think in terms of a wave function, the probability is given by its square, which neglects the phase. Usually it is more convenient to just see it as a state in Hilbert space.

To keep the debate going I should, however, mention that there are other interpretations. The Schrodinger equation can be re-written as a system with a classical part (trajectories) and one that depends on a global 'quantum potential'. This approach by Bohm isn't very popular at the moment, but I used to know someone who liked it.

 

[jar2574]

However, the point of the thought experiment fits neatly into your summary because it is meant to demonstrate that souls are irrelevant to creating intelligence either because they aren't necessary, or because there will inevitably be one in an AI that we create.

Your thought experiment did not demonstrate that souls are irrelevant to creating intelligence.

Souls could have no affect on the physical world, but be necessary for human intelligence to exist.

Your experiment could prove that souls are unnecessary, if it actually occurred and intelligence was created. But we cannot conclude that would necessarily be the result, because no one has proven that intelligence is solely the product of the physical world and no one has proven that souls don't exist.

Not only did you not understand the concept of a thought experiment, but you also did not understand the point I was making with it.

I understand the concept of a thought experiment. I did not understand your point, because the point you were attempting to make cannot be made using your experiment.

 

[Jimbo30]

Interesting topic. I didn't read it all obviously, just the first couple of pages.

I don't know if anyone else has stated this or not, but I'd like to make the following point...

If you could program an AI to play Civ4 like a human, it would probably find something better to do with it's time.

 

[warpus]

No one has proven that human intelligence is solely a product of the physical world. No one has proven that souls do not exist.

The burden of proof lies with the person who claims that souls exist.

For example, when Einstein formulated his theories of relativity, the burden of proof was his - he had to prove to the world that his theories were right. He did not say "Everything is relative to the observer" and then "Prove that I'm wrong".

The burden of proof always has to lie with the person making the claim.

If you claim that souls have something to do with intelligence - the burden of proof lies with you. If I claim that natural processes can create intelligence - the burden of proof lies with me.. not to prove that souls don't exist - but to prove what i'm claiming - namely that natural processes can produce intelligence.